The third theme from our Perth and Melbourne CFO roundtables was the one every finance leader recognised instantly: Copilot is already in place, it's genuinely useful — and it hasn't moved the needle on the business.
Almost every finance team we spoke to has Copilot rolled out. People use it to draft emails, summarise documents, get a first pass at a memo done faster. Nobody in the room disputed that it helps. What nobody could point to was a case where it changed how the finance function actually operates.
The reason came up repeatedly once we started digging: Copilot is a general-purpose assistant sitting on top of existing tools — Outlook, Word, Teams. It isn't wired into the systems where finance's real leverage lives: the ERP, the consolidation engine, the reporting stack, the reconciliation workflows that consume the bulk of a finance team's time each month. It can help someone write a better summary of a report. It can't pull from the general ledger, reconcile against source data, or validate an output against the controls finance already runs everything else through.
The gap between "help me write this" and "help me close the books"
That's the ceiling. Individual productivity tools sit at the edge of the finance function — the email, the memo, the summary — because that's as far as a general-purpose assistant can reach without deep integration into finance's actual systems and data. Enterprise productivity sits at the centre: month-end close, consolidation, reporting cycles, forecasting. None of that gets faster because one person's inbox got easier to manage.
Every finance exec in the room had lived some version of this. Copilot proved AI could help. It didn't show them how to get AI genuinely inside the processes that drive the bulk of the function's time and cost — because that requires integration work Copilot was never built to do.
Copilot's ceiling is a temporary problem, not a permanent one. Every major finance platform — ERP, consolidation, planning, reporting — is already moving to embed its own native AI agents directly into the system, rather than leaving finance to bolt on a general-purpose assistant from outside. That's the integration Copilot couldn't offer, arriving from the other direction: the tools finance already runs on are about to come with agents built in.
We think this moves faster than most finance teams expect. Within a couple of years, we expect finance functions to have more AI agents operating inside their systems than actual humans — one agent reconciling, one drafting forecasts, one flagging anomalies, one preparing audit evidence, each embedded in a different tool.
That changes the question entirely. It stops being "does our productivity tool reach far enough into the process" and becomes "who is governing and coordinating a growing population of agents across our systems." The organisations that get real enterprise value won't be the ones with the most agents — they'll be the ones that can govern, harness, and orchestrate them as a coherent whole: consistent controls, clear ownership, and a single view of what every agent is doing and why. Without that, more agents just means more ungoverned surface area for audit to worry about.
This is exactly the gap our workshops are built to close. Rather than starting with a model or a licence, we sit down with the client's actual setup — their systems, their processes, their data — and work through where AI can genuinely reach into the workflow, not just sit alongside it. It's a collective exercise: the finance team knows the process end to end, we know how to identify where integrated AI can realistically change it, and together we prioritise the use cases with the clearest path to enterprise-level impact.
Increasingly, that also means designing the governance layer alongside the use cases — not as an afterthought, but as the thing that lets a finance team actually harness a growing set of agents rather than just accumulate them. The output isn't a list of everything AI could theoretically touch. It's a short, ranked set of use cases, plus a clear view of who owns and governs each agent involved, so the finance team can scale up with control rather than sprawl.
Every CFO in the room agreed on two things: AI in finance is here to stay, and the number of agents doing finance work is only going up from here. The harder question is getting past general-purpose tools sitting at the edges of the function, into a position where finance can genuinely govern and orchestrate the agents now embedding themselves in every system it runs. That takes deliberate integration and governance work, not another licence rollout — and it's the conversation we want to keep having.
Join the conversation. We're bringing this discussion to Sydney next. If your finance team has Copilot but hasn't found a way past individual productivity, come work through it with us. Register here to join the session.